I don't know if officially there are devices with mali 4xx, but I solved it by applying this patch to mesa 21.0.3.
this patch is now also upstream.
Looks like it's bcm4329.
Codeh96miniv8:/ # cat /sys/firmware/devicetree/base/dwmmc@30010000/wifi@1/compatible brcm,bcm4329-fmacEDIT: this seems to be wrong - it's coming straight from the dts. Is it possible to get the wifi card model from the sysfs, or do I have to open up the box?
unfortunately yes, you have to open the box and read the label of the wifi chip, or if you have another working SO (like Android), retrieve the information in another way.
ilovefly if you have a spare sd card why not 
but keep in mind that often info about ram can be faked on Android (on libreelec instead it's the real one)
I updated mesa to the latest version and saw that there are several improvements and bugfixes for lima driver.
the problem seems to have disappeared
cheers
Michele
Hi everyone,
I'm working on an unofficial libreelec 10 build for soc rockchip 322x (not released yet).
It's a mix of the work done by knaerzche and some patches taken by armbian ( fabiobassa and jock2)
It works pretty well but I only have one issue that I don't know if I have solved it well.
Starting streams decoded via software (netflix addons and similar) after about 6 minutes there is a kodi crash due to an EGL_BAD_ALLOC error (log attached).
I tried with different cma memory sizes (from 64MB to 512MB) and also lower resolutions, but the only solution was to disable the dmabuf management in the kernel.
I suspect a memory leak but I'm not sure. Can anyone help me!?
Cheers
Michele
bluetako wich builds?! mainline or legacy?!
Can you please share more info of WiFi chip?! (brand and model)
cheers
hi chewitt
to obtain this result we studied the flow a lot and understood that, for example, the driver used a custom implementation for the encryption protocols, when in reality the chip can do it at the hardware level/using the kernel routines.
Therefore a lot of code incompatible with the mainline kernel has been removed and porting has become easier.
But we worked on what we had (rockchip) so the custom code for amlogic was removed because we couldn't test it.
The original driver has different behaviors for amlogic, and the error you receive is related to how the mac address is retrieved (rockchip read efuse .. amlogic seems to use static uboot parameter)
Clearly having an amlogic box would be more useful, and I thank you for the proposal.
But I do it in my little free time and I would not like to promise something that I cannot keep.
Hi all,
with fabiobassa and jock2 we are experimenting with the 6051p driver.
The idea is to rewrite it, but in the meantime we have managed (by cleaning up the code a bit) to have a working version for the mainline kernels.
It needs firmware (ssv6051-sw.bin) and a configuration file (ssv6051-wifi.cfg) under /lib/firmware in order to work properly.
Don't tested yet with libreelec (my working on mainlining libreelec for rk322x is not yet usable), but with armbian seems to works well.
markosc, I've updated only the image for v884k and v88mars.. try and let me know
I just use it on an sd and compared to the previous build I see that I can't connect via ssh and samba files and I'm connected via wifi which looks powerful. my tv box is rk3229 V88 4k...Cheer
Thanks for the report. I'll try to fix it in the next build.
My images contains some optimizations, and works better, but rk322x is an old soc.
So what i meant is that to try to improve even more, i should try to overclock gpu and cpu.
and I don't know if it makes sense to risk system stability for a gui.
for the usb problem you could try to copy the dtb file of the image that works, and use it with my images.
share kernel logs anyway. I'll try to take a look.
thanks
sadly rk322x is not rpi4
so fast UI at 4k with all visual effects enabled is impossible.
I suggest to stay at 720p for a better UI experience.
in the meantime i managed to build an alpha version of libreelec 10.. it works quite well ... but please don't ask ETA
Yes. However, I'm not sure that all the things that are working with legacy kernel will work also in mainline kernel (NAND for example).
This is because the legacy kernel is the official one from rockchip, instead, the mainline kernel works thanks to the work of volunteers
Hi,
last update for my libreelec-9.2 fork:
- fixed rkmpp memory leaks
- small improvements
as usual I've builted all the images.
Hi shMatrix, to optimize even further, in the next build I've disabled some kernel options used by developers (aka debugfs).
This can improve something, but I don't expect miracles.
My suggestion is to try with another image considering that the one for mxq4kpro is for rk3228a. This soc in teory should have a lower clock for cpu and gpu (see first post for details).
Often these boxes are sold as a model, but internally they may be different.
For example mine is an mxq4k, but the board is identical to the one present in v88 mars
So check in the first post if there is an image that is similar to the one you use(ddr3, same wifi chip, same internal memory etc etc), but suitable for rk3229 / rk3228b and download it from my gdrive.
OK thanks. For now i will use a micro sd card just in case. Is the builds from 6th Jan this year the best to use?
try it and let me know. I'm planning another build (there is a bug for sure) but I prefer to collect all the wrong things before 
Hi ushilives, you can install on NAND in many ways:
- use rkdevelop tool (hard)
- dd from a working libreelec (medium)
- use the great multitool from armbian port (easy)
Installing on NAND should make the boot a bit faster, but the major issue is that in order to update the image you need to upload it manually via ssh (or burn again).
This is because I didn't setup yet an update site for my builds.
p.s.obviusly, if you install libreelec on NAND, you will lost the stock firmware of your box.
thanks Dave, seems a bug of rkmpp.
I can try to fix, but can you please tell me which image are you using in order to rebuild only that one for testing?!
cheers
